fix: handle null response.output in parse_response

[xlyoung]

Problem

parse_response() in src/openai/lib/_parsing/_responses.py iterates response.output without checking for None. The chatgpt.com Codex backend sometimes sends response.output: null in the consolidated response.completed event, even when valid response.output_item.done events were streamed earlier in the same response.

This causes TypeError: 'NoneType' object is not iterable, killing the entire stream before the consumer can read the deltas.

Fix

Use (response.output or []) to handle null output gracefully:

# Before
for output in response.output:

# After  
for output in (response.output or []):

Testing

• Import test passes
• No existing tests for parse_response to verify against
• The fix is a defensive null check that doesn't change behavior when output is present

Fixes #3325

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4a9b095c0a

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve streamed output when completed payload is null

When the streaming backend sends response.completed.response.output: null after already sending response.output_item.* and text/function deltas, this turns the completed response into output=[]. In ResponseStreamState.accumulate_event, the final response.completed event and get_final_response() are built from parse_response(event.response), not from the accumulated snapshot, so consumers no longer crash but silently lose all streamed output and parsed content in exactly the scenario this change is meant to handle. The null case should fall back to the accumulated snapshot/output rather than an empty list.

Useful? React with 👍 / 👎.